Endangered Biodiversity

Wildlife Tracking and Modeling of Endangered Animals

Pantanal is a rich ecosystem that houses many wildlife animals, including reptile, amphibian, bird and mammal species. However, with the expansion of agriculture, the Pantanal biome needs to conciliate the demand for rural development while conserving biodiversity. Similarly, areas of power plant dams have intentionally flooded big portions of land, intending to generate more power but affecting the environment. Thus, it is urgent to track and estimate the population to understand the impact of the power industry and agricultural activities on Pantanal's biodiversity. This project aims to automatically identify endangered animals and help ecologists and biologists manage the population of animals in wetlands.

Image Acquisition using Thermal and Visual Cameras

Images Captured with Camera Traps

Images from Small Unmanned Aerial Vehicles (UAVs)

Biodiversity Tracking and Sustainable Production

Motivation: Track the local biodiversity and curb deforestation are key for a susitainable production. Camera traps are used as remotely activated devices equipped with a motion sensor. Typically, they are fixed in trees to monitor wildlife animals.

Gap: It is very complex for researchers to set camera traps in wetlands, including power plant flooded areas in Sao Paulo and the biome Pantanal where concentrated portions of land and small vegetation are covered with water.

Solution: Besides camera traps, UAVs are suitable tools to capture images and videos of animals. To this end, we can adapt different thermal and visual cameras and create specific flight plans to analyze the entire area. Deep learning-based methods were developed to detect and classify animals with both spectrums.

Publications of the Project

Thermal Image Segmentation in Studies of Wildlife Animals

Workshop of Computer Vision, 2015
Mauro Santos et al.

Recognition of Pantanal Animal Species using Convolutional Neural Networks

Workshop of Computer Vision, 2016
Diogo Gonçalves et al.

Recognition of Endangered Pantanal Animal Species using Deep Learning Methods

International Joint Conference on Neural Networks, 2018
Mauro Santos et al.

Who we are:

We are a multidisciplinary team composed by Computer Engineers, Biologists and Ecologists. We had been interested in creating solutions for automatically identifying wild species from big data sets. The researchers of this project are:

  • Bruno Brandoli
  • Wesley Nunes Gonçalves
  • Mauro Santos de Arruda
  • Diogo Nunes Gonçalves
  • Gabriel Spadón
  • Jose F Rodrigues-Jr
  • Marcia Gomes
  • Walfrido Moraes Tomas